Properties of inverse-based controllers under failure conditions are addressed. The inverse-based controllers considered are parameter&d, designed and implemented under the internal model control (IMC) structure. One important property is that the integral action is lost when failure in actuator and/or sensor occurs. The steady-state errors (offsets) under failure conditions are discussed in detail from aspects of the set-point tracking and disturbance rejection. The steady-state errors for set-point tracking can be characterized by the well-known interaction measures such as RCA and ERG. The overall performance under failure status is assessed based on the sum of squares of steady-state errors (SSE). An amplification factor (AF) is defined as the ratio of SSE with IMC control over the SSE without control. The results show that, under failure modes, AF can be greater than one in some cases. This implies IMC feedback system may lead to inferior performance than a system without control. A distillation example is used to illustrate the behavior of IMC system under failure modes. The results presented here provide a better understanding of IMC and give useful information in making on-line decision under failure conditions. 'Author to whom correspondence should be addressed. plementation between these two structures can lead to significantly different properties under failure conditions. In chemical process control, multivariable controllers are often designed to eliminate steady-state offsets. The properties of inverse-based multivariable controllers under failure conditions with conventional feedback are studied by Chang and Yu (1991). The results show that, under failure condition, i.e. a combination of sensor and actuator failure, we are able to maintain zero steady-state offsets for the remaining controlled variables. However, instability may occur under failure conditions as a result of positive feedback. Since the integral action for the inverse-based controller under IMC structure arises implicitly, under failure conditions, IMC behaves differently. The purpose of this work is to investigate the performance of IMC under actuator failure, sensor failure and a combination of actuator and sensor failure.